A deep neural network integrating multiple echocardiography video views significantly improves cardiac disease detection over single-view models.
Key Details
- 1Researchers at UCSF developed a multiview deep neural network (DNN) architecture for echocardiography.
- 2The system integrates information from three standard echo views (A4c, A2c, PLAX) using a midfusion approach.
- 3Compared to single-view networks, the multiview DNN improved AUC by 0.06 to 0.09 across tasks.
- 4On UCSF test data, AUCs were 0.91 (LV/RV abnormality), 0.84 (diastolic dysfunction), and 0.90 (valve regurgitation), with sensitivities and specificities ranging from 74%-84%.
- 5External validation (Montreal Heart Institute): AUCs were 0.91, 0.92, and 0.79 for the three tasks.
- 6Authors stress the potential of multiview AI for broader medical imaging applications.
Why It Matters
This work provides evidence that AI models simulating radiologists' multi-view interpretative strategies can match or surpass conventional single-view approaches, suggesting a path toward more accurate and generalizable automated disease detection in echo and potentially other imaging modalities.
Source
AuntMinnie
Related News
•AuntMinnie
Study: Computer Vision Models Best LLMs in Chest CT Breast Abnormality Detection
Computer vision models (CVMs) surpass large language models (LLMs) in accurately labeling incidental breast abnormalities on chest CT scans.
•Cardiovascular Business
Radiology Maintains Lead in FDA-Cleared AI Algorithms, Cardiology Follows
Radiology remains the top specialty for FDA-cleared AI, with cardiology as a strong second, particularly in cardiovascular imaging.
•AuntMinnie
Deep Learning Models Rival Radiologists for Pancreatic Cancer Detection on CT
Deep-learning models achieved comparable or superior accuracy to experienced radiologists in detecting pancreatic cancer on CT scans, especially for small tumors.